Trials and tribulations: Using patients\u27 and physicians\u27 preferences to enhance the efficiency and ethics of randomized clinical trials

The concomitant problems of underenrollment and selective enrollment limit the efficiency of many randomized clinical trials (RCTs). Because underenrollment reduces precision, and selective enrollment reduces generalizability, ensuing trials may have limited value. This shifts the risk/benefit ratio in an unfavorable direction for the many patients who consider enrolling in trials for altruistic reasons. Past efforts to overcome these problems are not broadly applicable, and many entail their own methodological and ethical problems. This dissertation describes a new method, called prospective preference assessment (PPA), that may simultaneously enhance the efficiency and ethics of RCTs. The foundation of participant PPA rests upon the ability of prospective research participants\u27 stated preferences for trial enrollment to predict their actual enrollment decisions; clinician PPA requires that clinicians\u27 stated preferences for referring their patients for enrollment correspond to their actual referral behaviors. We tested the first of these assumptions, and found that among 499 people at high risk for HIV infection, stated willingness to participate in a hypothetical AIDS vaccine trial was the single best predictor of actual enrollment decisions, albeit with modest properties of discrimination and calibration. We next evaluated the ability of participant PPA to elicit 126 hypertensive patients\u27 preferences for enrolling in placebo-controlled trials of new antihypertensive drugs. We found that increasing probabilities of being assigned to a placebo group or experiencing an adverse drug effect discouraged some patients from participating, but were unlikely to contribute substantially to underenrollment or selective enrollment. We also found that financial inducements encouraged enrollment; however, such inducements did not meet predefined criteria for being undue or unjust. We evaluated the ability of clinician PPA to determine physicians\u27 preferences for active-controlled versus placebo-controlled trials of antihypertensive drugs. We found that physicians reported being significantly more willing to refer their patients to, and use information from active-controlled trials than placebo-controlled trials. Because mailed physician surveys are often limited by low response rates, we also conducted a randomized trial to evaluate three strategies to improve response rates. We found that larger financial incentives represent an effective and cost-effective approach, but that neither envelope size nor the provision of mint candies influenced response rate. Finally, we discuss how prospective measures of patients\u27 treatment preferences can be incorporated into analyses of RCT data to provide less biased estimations of the specific effects of interventions. We conclude that PPA is a feasible method to elicit both patients\u27 and physicians\u27 enrollment preferences. The method may help future investigators design trials that respect patients\u27 interests, may help determine patterns of selective enrollment, and may allow the more valid analysis of RCT results. However, because patients\u27 enrollment preferences often depend on immutable characteristics of trials, PPA may have only modest effects on participant accrual